Integrating Predictive Visualization with the Epidemic Disease Simulation System (EpiSimS)

The Epidemic Simulation System (EpiSimS) is a scalable, complex model for analyzing disease spread within the United States. Due to the high-dimensional parameter space, the long completion time, and the large amount of output data in a single EpiSimS run, simulating the entire input parameter space is unfeasible. By taking a granular sampling of the parameter and aggregate outcome space, regression algorithms can predict outcomes that a particular parameter combination lead to, without having to actually run the simulation. These predictions are viewable using traditional epidemic visualization components: aggregate line charts and spatio-temporal mapping. Our ongoing effort is to integrate predictive algorithms into our base EpiSimS viewer, where a user can load completed runs to build a classifier, choose an arbitrary input parameter set, and see the predicted outcomes using visual means. Our future work involves developing faster and more accurate prediction algorithms into our system and leveraging prediction to enable specific location classification and response countermeasures.